Empirical Rule Calculator — 68-95-99.7 Rule
Show results for:
What Is the Empirical Rule?
The empirical rule is a shorthand description of how data spreads around the mean in an approximately normal distribution. It tells you, without any complex calculation, that the vast majority of observations cluster close to the average and that the proportion doing so follows a predictable pattern.
Specifically, for a distribution that is approximately normal (bell-shaped and roughly symmetric), three statements hold:
About 68% of observations fall within 1 standard deviation of the mean (μ ± σ).
About 95% fall within 2 standard deviations (μ ± 2σ).
About 99.7% fall within 3 standard deviations (μ ± 3σ).
The rule is also called the 68-95-99.7 rule, a name that makes its three percentages impossible to forget. It was formally described in work connecting the normal distribution to practical data analysis, and it remains one of the first tools statisticians reach for when making quick judgments about a dataset.
The 68-95-99.7 Rule
Each of the three percentages in the rule describes a distinct band of data centered on the mean. They are cumulative and nested: the 95% region contains the 68% region, and the 99.7% region contains both.
| Interval | Formula | % Inside | % Outside | % Per Tail |
|---|---|---|---|---|
| μ ± 1σ | μ − σ to μ + σ | ~68% | ~32% | ~16% each |
| μ ± 2σ | μ − 2σ to μ + 2σ | ~95% | ~5% | ~2.5% each |
| μ ± 3σ | μ − 3σ to μ + 3σ | ~99.7% | ~0.3% | ~0.15% each |
Reading across the table, the intervals widen as k increases, and the percentage inside grows closer to 100%. The percentage outside shrinks rapidly: from 32% beyond 1σ, to 5% beyond 2σ, to just 0.3% beyond 3σ. This rapid thinning of the tails is a fundamental property of the normal distribution.
Percentage Breakdown Between Boundaries
It is often useful to know the percentage of data that falls between two standard-deviation boundaries, not just within a single centered band. These values follow directly from the nested structure of the intervals.
| Region | Approximate % |
|---|---|
| Within ±1σ (center) | ~68% |
| Between 1σ and 2σ (each side) | ~13.5% |
| Between 2σ and 3σ (each side) | ~2.35% |
| Beyond ±3σ (each tail) | ~0.15% |
Verification: the full distribution must sum to 100%. The center holds 68%, leaving 32% split evenly as 16% in each tail. The 95% band holds 95%, so the region between 1σ and 2σ on each side is 95/2 − 68/2 = 47.5 − 34 = 13.5%. The 99.7% band gives 99.7/2 − 95/2 = 49.85 − 47.5 = 2.35% between 2σ and 3σ on each side.
Empirical Rule Formula
The general formula for any empirical rule interval is:
Lower Bound
Lower = μ − kσ
Upper Bound
Upper = μ + kσ
Central Interval
(μ − kσ, μ + kσ)
Variables
μ = mean
σ = standard deviation
k = 1, 2, or 3
Set k = 1 to get the 68% interval, k = 2 for the 95% interval, and k = 3 for the 99.7% interval. The mean is always the midpoint of the resulting range, and the interval extends kσ units on either side.
Worked Examples
Example 1: IQ Scores (μ = 100, σ = 15)
IQ scores are standardized so that the population mean is 100 and the standard deviation is 15. Assuming the distribution is approximately normal:
Lower = 100 − 15 = 85
Upper = 100 + 15 = 115
About 68% of IQ scores fall between 85 and 115.
Lower = 100 − 30 = 70
Upper = 100 + 30 = 130
About 95% of IQ scores fall between 70 and 130.
Lower = 100 − 45 = 55
Upper = 100 + 45 = 145
About 99.7% of IQ scores fall between 55 and 145.
Interpretation: Only about 0.3% of IQ scores lie outside the range 55 to 145. An IQ below 55 or above 145 places an individual in approximately the bottom or top 0.15% of the population.
Example 2: Finding a 95% Interval (μ = 75, σ = 8)
You want to find the range that contains approximately 95% of the data for a distribution with mean 75 and standard deviation 8.
Lower = 75 − 2(8) = 75 − 16 = 59
Upper = 75 + 2(8) = 75 + 16 = 91
The 95% interval is 59 to 91.
Example 3: Expected Counts (μ = 50, σ = 5, n = 2,000)
A factory produces components with a mean diameter of 50 mm and a standard deviation of 5 mm. Quality control measures 2,000 components. How many are expected to fall within each interval?
Within 2σ (40 to 60 mm): 0.95 × 2,000 = ~1,900 components
Within 3σ (35 to 65 mm): 0.997 × 2,000 = ~1,994 components
These are expected counts under the empirical rule, assuming the diameter distribution is approximately normal. Actual counts may vary.
Example 4: Percentage Outside an Interval
What percentage of data lies outside 2 standard deviations of the mean?
Since the normal distribution is symmetric, approximately 2.5% lies below μ − 2σ and approximately 2.5% lies above μ + 2σ.
Example 5: Percentage Between the Mean and a Boundary
What percentage falls between the mean and the point 1 standard deviation above it?
Answer: approximately 34% lies between μ and μ + 1σ.
Empirical Rule vs Exact Normal Probabilities
The percentages in the empirical rule are rounded approximations of the exact values produced by the standard normal distribution's cumulative distribution function. This distinction matters whenever you need precise probability calculations.
| Interval | Empirical Rule | Exact Normal Probability |
|---|---|---|
| μ ± 1σ | 68% | 68.27% |
| μ ± 2σ | 95% | 95.45% |
| μ ± 3σ | 99.7% | 99.73% |
For most practical purposes in introductory statistics and quick analysis, the rounded values are adequate. If you need the precise probability for any z-score or interval, use the Normal Distribution Calculator or a Z-table instead.
Empirical Rule vs Chebyshev's Theorem
Both the empirical rule and Chebyshev's theorem describe how data clusters around the mean, but they serve different situations and make different assumptions.
| Feature | Empirical Rule | Chebyshev's Theorem |
|---|---|---|
| Distribution required | Approximately normal | Any distribution with finite variance |
| k = 1 | ~68% | No useful bound (k must exceed 1) |
| k = 2 | ~95% | At least 75% |
| k = 3 | ~99.7% | At least 88.9% |
| Result type | Specific approximation | Lower bound (at least) |
| Accuracy | Precise for normal data | Conservative for all data |
Chebyshev's theorem guarantees a minimum percentage regardless of distribution shape. The empirical rule gives a close estimate for approximately normal data. When data is clearly non-normal, use Chebyshev's theorem. When data follows a bell curve, the empirical rule is more informative.
Explore both with the Chebyshev's Theorem Calculator.
Empirical Rule and Z-Scores
The empirical rule maps directly onto the z-score framework. A z-score measures how many standard deviations a value lies from the mean:
Z-Score Formula
z = (x − μ) / σ
A value 1 standard deviation above the mean has z = 1. Two standard deviations below the mean gives z = −2. The empirical rule boundaries therefore correspond to z = ±1, z = ±2, and z = ±3 on the standard normal distribution.
So the statement "about 68% of data falls within 1 standard deviation" is equivalent to saying "about 68% of observations have a z-score between −1 and +1." Use the Z-Score Calculator to find the z-score for any specific observation.
When to Use the Empirical Rule
The empirical rule is appropriate when the dataset is:
- Approximately bell-shaped and symmetric around the mean
- Unimodal (one clear peak), not bimodal or multimodal
- Free of extreme outliers that distort the tails
- Drawn from a process well-modeled by the normal distribution
Common datasets where the rule works well include physical measurements (height, weight), standardized test scores, manufacturing tolerances, and many natural phenomena. Checking whether your data is approximately normal first is good practice. A bell curve diagram or a Q-Q plot can help assess normality visually.
When Not to Use the Empirical Rule
Avoid the empirical rule when data shows any of the following characteristics:
- Strong skewness: income distributions, wait times, and many count-based measurements are right-skewed
- Heavy tails: financial returns often have more extreme values than a normal model predicts
- Multiple modes: a bimodal dataset is not normal and the rule will give misleading percentages
- Substantial outliers: a few extreme values can make the standard deviation misleading
- Bounded data: proportions and percentages that are bounded between 0 and 1 are rarely well-approximated by a normal model
Common Empirical Rule Mistakes
How to Use This Calculator
Type any numeric value. The mean can be positive, zero, or negative.
Must be a positive number greater than zero. If you need to calculate standard deviation first, use the Standard Deviation Calculator.
If you provide n, the calculator shows the approximate number of observations expected within each interval. Leave blank for percentage-only results.
Use the selector to show all three intervals at once, or focus on 1, 2, or 3 standard deviations.
Results appear immediately: lower and upper bounds, inside and outside percentages, tail percentages, and a normal distribution diagram.
Confirm that your data is approximately normally distributed before drawing conclusions from the results.
Related Tools and Topics
The empirical rule connects to several core concepts in statistics. These pages cover the underlying theory in more depth.
Frequently Asked Questions
The empirical rule states that for an approximately normal distribution, about 68% of observations fall within 1 standard deviation of the mean, about 95% fall within 2 standard deviations, and about 99.7% fall within 3 standard deviations. It is also called the 68-95-99.7 rule.
About 68% of observations fall within 1 standard deviation of the mean in an approximately normal distribution. This means the interval from μ − σ to μ + σ contains roughly 68% of the data. The remaining 32% lies outside this range, split approximately evenly as 16% in each tail.
Use the formula Lower = μ − kσ and Upper = μ + kσ, where μ is the mean, σ is the standard deviation, and k is 1, 2, or 3. For example, with μ = 100 and σ = 15, the 95% interval (k = 2) is Lower = 100 − 30 = 70 and Upper = 100 + 30 = 130.
Approximately 5% lies outside 2 standard deviations (100% − 95% = 5%). Because the normal distribution is symmetric, about 2.5% lies below μ − 2σ and about 2.5% lies above μ + 2σ.
No. The empirical rule only applies to distributions that are approximately normal: bell-shaped, roughly symmetric, and unimodal. It should not be applied to strongly skewed, heavy-tailed, or multimodal data. For a rule that applies to any distribution, use Chebyshev's theorem instead.
No. The exact probability within ±1 standard deviation in a normal distribution is approximately 68.27%, not exactly 68%. The empirical rule rounds this for convenience. Similarly, 95% rounds 95.45%, and 99.7% rounds 99.73%. For precise probability calculations, use a normal distribution table or calculator.
The empirical rule requires an approximately normal distribution and gives specific estimates (68%, 95%, 99.7%). Chebyshev's theorem applies to any distribution with finite variance and provides a lower bound: at least 1 − 1/k² of data lies within k standard deviations. For k = 2, Chebyshev guarantees at least 75%, while the empirical rule says approximately 95% for normal data.
Multiply the proportion by the sample size: expected count = percentage × n. For n = 500 observations with an approximately normal distribution, you would expect 0.68 × 500 = 340 observations within 1 standard deviation, 0.95 × 500 = 475 within 2 standard deviations, and 0.997 × 500 = 498 or 499 within 3 standard deviations. These are approximate expected values, not guarantees.
The empirical rule boundaries correspond directly to z-score values. The 68% interval covers z-scores from −1 to +1. The 95% interval covers z-scores from −2 to +2. The 99.7% interval covers z-scores from −3 to +3. A z-score of 1 means the observation is exactly 1 standard deviation above the mean, placing it on the outer edge of the 68% band.
It means that for an approximately normal distribution, nearly all observations (about 997 out of every 1,000) fall within 3 standard deviations of the mean. Only about 3 in 1,000 fall outside this range. In quality control, the "3-sigma rule" uses this property to flag production processes: if a measurement falls beyond 3 standard deviations, it is a statistically unusual event worth investigating.